A rare disease recognition system based on neural network

Abstract

The present invention relates to the field of image recognition technology, and particularly to a rare disease recognition system based on a neural network. The system includes a sample collection module, a sample screening module, a sample scanning module, a sample recognition module, and a sample evaluation module. The present invention improves the efficiency and consistency of data collection by automatically collecting rare disease samples. The sample screening module preprocesses and screens the original data set to improve the efficiency of data processing and ensure data quality. The sample scanning module accurately segments the lesion area in the scanned image through image recognition technology, providing an accurate basis for subsequent analysis. The sample evaluation module calculates the real-time accuracy rate to dynamically adjust the screening parameters according to the real-time accuracy rate, improving the accuracy of recognition, enabling the model to perform refined learning for the target task, enhancing the applicability of the model, and at the same time reducing the dependence on large-scale labeled data and saving the computing power resources required for training the deep learning model.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and in particular, to a rare disease recognition system based on a neural network. Background Art

[0002] There are more than 6,000 rare diseases globally, affecting more than 300 million people. Early accurate diagnosis has always been an important goal in the clinical practice of rare diseases. Among them, about 80% are related to genetic factors. Rare diseases are difficult to diagnose and the time taken to confirm a diagnosis is long. Accurately identifying the lesion area of rare diseases is a great challenge. Nowadays, computer vision has developed rapidly. How to introduce computer vision technology into the work of detecting medical images to improve the efficiency and accuracy of image diagnosis. In computer vision recognition methods, the convolutional neural network (CNN) is a deep learning model particularly suitable for processing image data. CNN consists of multiple convolutional layers and pooling layers, can automatically learn the hierarchical feature representation of images, and has powerful expressive ability and generalization ability. It has achieved great success in the fields of image classification, object detection, semantic segmentation, etc., and is also widely used in image analysis.

[0003] The patent document with the publication number CN112472048A discloses a neural network structure for pulse condition recognition of cardiovascular disease patients, including the following steps: S1, pulse condition data collection; S2, pulse condition data preprocessing; S3, establishing the neural network structure; S4, training process. Thus, in the existing disease detection technology, there is a lack of a recognition system suitable for rare lung diseases, which can accurately and quickly screen the input image data set, so as to input high-quality images into the CNN model to accurately intercept the rough segmentation area of the lesion area and improve the recognition efficiency. Summary of the Invention

[0004] Therefore, the present invention provides a rare disease recognition system based on a neural network to overcome the problem in the prior art that due to the lack of rapid and accurate screening of a large number of input data sets, the image quality input into the CNN model is low, resulting in low recognition efficiency.

[0005] To achieve the above object, the present invention provides a rare disease recognition system based on a neural network, including,

[0006] A sample collection module, which is connected to an external big data platform and is used to collect rare disease samples at a preset collection period to obtain an original data set;

[0007] A sample screening module, which preprocesses and screens the original data set in sequence to obtain a target data set;

[0008] A sample scanning module is used to obtain each scanned image in the target dataset, determine the real-time line area ratio according to the first standard line area ratio and the second standard line area ratio, select a processing mode according to the judgment result, and identify and segment the corresponding lesion areas in each scanned image according to the processing mode and a convolutional neural network to obtain a roughly segmented image;

[0009] A sample recognition module is used to recognize the roughly segmented image, extract image features, compare the image features with lesion features, and mark the matching roughly segmented images as target image samples;

[0010] A sample evaluation module calculates the percentage of the number of target image samples correctly predicted by the model in the total image samples to obtain the real-time accuracy rate, so as to determine whether to adjust the screening parameters;

[0011] Among them, the screening parameter is the second standard line area ratio.

[0012] Further, the sample screening module includes a preprocessing unit, a primary screening unit, and an iterative screening unit, where

[0013] The preprocessing unit is used to obtain the image data in the original dataset as the original image, perform contour enhancement processing on the original image to obtain an intermediate image;

[0014] The primary screening unit performs a primary screening on each intermediate image to delete the intermediate images with unqualified quality to obtain an intermediate dataset;

[0015] The iterative screening unit performs continuous extended scanning on the intermediate images in the intermediate dataset to obtain the target dataset.

[0016] Further, the sample scanning module includes a window sliding unit, an analysis unit, a selection processing unit, and a segmentation unit, where

[0017] The window sliding unit can obtain the total number of pixels of any scanned image, calculate the preset sliding window size according to the sliding scanning coefficient and the total number of pixels, divide the scanned image into several areas to be scanned with the preset sliding window size, and perform sliding scanning on the areas to be scanned;

[0018] The analysis unit is used to obtain the marked line pixels in the area to be scanned, and calculate the unit line pixel ratio and the real-time line area ratio according to the marked line pixels;

[0019] The selection processing unit is used to select a processing mode according to the judgment result, so as to identify and segment the corresponding lesion areas in each scanned image according to each processing mode and a convolutional neural network to obtain a roughly segmented image;

[0020] The segmentation unit is used to segment the image to be analyzed and the area to be scanned to obtain a roughly segmented image.

[0021] Further, the analysis unit includes a determination subunit, a marking subunit, and a calculation subunit, where

[0022] The determination subunit is used to compare the proportion of unit line pixels with the standard proportion of line pixels;

[0023] The marking subunit is used to mark the image area corresponding to the current sliding window as a line area when the proportion of unit line pixels is greater than or equal to the standard proportion of line pixels;

[0024] The calculation subunit is used to obtain the percentage of the number of line areas in the total number of areas to be scanned to obtain the real-time proportion of line areas.

[0025] Further, the selection and processing unit includes a comparison subunit and a processing subunit, where

[0026] The comparison subunit is used to compare the real-time proportion of line areas according to the first standard proportion of line areas and the second standard proportion of line areas;

[0027] The processing subunit is used to enhance the line contour edge of the area to be scanned when the real-time proportion of line areas is between the first standard proportion of line areas and the second standard proportion of line areas to obtain the image to be analyzed.

[0028] Further, the segmentation unit includes a first segmentation subunit, a second segmentation subunit, and a third segmentation subunit, where

[0029] The first segmentation subunit, when the real-time proportion of line areas is less than the first standard proportion of line areas, performs nodule feature recognition and segmentation on the area to be scanned according to a convolutional neural network to obtain a roughly segmented image;

[0030] Among them, the first segmentation subunit identifies nodule features according to a convolutional neural network and segments the area where the nodule features are located to obtain a roughly segmented image;

[0031] The second segmentation subunit is used to identify the image to be analyzed to obtain a roughly segmented image;

[0032] The third segmentation subunit is used to perform recognition and segmentation on the area to be scanned according to a convolutional neural network to identify non-nodular pulmonary lesion areas to obtain a roughly segmented image when the real-time proportion of line areas is greater than the second standard proportion of line areas.

[0033] Further, the processing subunit performs radial gradient transition processing on each gradient transition layer of the line contour edge to strengthen the line contour edge.

[0034] Further, the preliminary screening unit includes a data deletion subunit and a data enhancement subunit, where

[0035] the data deletion subunit is used to determine whether to delete the intermediate image according to the slice thickness;

[0036] the data enhancement subunit is used to perform contour edge marking on the edge pixel points of the picture area in the intermediate image area, and sequentially adjust each original brightness value to the corresponding corrected brightness value to perform data enhancement processing on the brightness data of the intermediate image.

[0037] Further, the data enhancement subunit respectively obtains the brightness values of each pixel point in the area within the contour as a number of original brightness values, calculates the average brightness within the contour according to each original brightness value, obtains the brightness values of each pixel in the area outside the contour to calculate the average brightness outside the contour, subtracts the average brightness within the contour from the average brightness outside the contour to obtain the real-time brightness difference, and sequentially adds each original brightness value to the real-time brightness difference to obtain the corresponding corrected brightness value to perform data enhancement processing on the brightness data of the intermediate image.

[0038] Further, the sample evaluation module includes a comparison unit and a correction unit, where

[0039] a accuracy rate threshold is set in the comparison unit, and the comparison unit compares the real-time accuracy rate with the accuracy rate threshold to obtain a comparison result;

[0040] the correction unit obtains the comparison result, and when the real-time accuracy rate is less than the accuracy rate threshold, corrects the proportion of the second standard line area to the proportion of the second corrected line area.

[0041] Compared with the prior art, the beneficial effects of the present invention are as follows: by connecting to a big data platform, rare disease samples are automatically collected, improving the efficiency and consistency of data collection. Through the sample screening module, the original data set is preprocessed and screened, improving the efficiency of data processing and ensuring data quality. The sample scanning module accurately segments the lesion area in the scanned image through image recognition technology, providing an accurate basis for subsequent analysis. Through the sample evaluation module, the real-time accuracy rate is calculated to dynamically adjust the screening parameters according to the real-time accuracy rate, optimizing the recognition process and improving the accuracy of recognition.

[0042] Further, by determining the slice thickness of the image, images with a slice thickness meeting the standard are automatically screened to ensure sufficient detail capture.

[0043] Further, by setting a scanning coefficient, the corresponding preset sliding window size is calculated according to the actual size of the image to be processed, improving the applicability of image processing.

[0044] Further, when it is determined that the proportion of the real-time line area is between the proportion of the first standard line area and the proportion of the second standard line area, there may be some blurred line edges. Since the blurred or unclear edges are also related to the lesion, the line contour edges are first enhanced to ensure the clarity of line recognition and avoid misrecognition. After enhancement, the line edge pixels are analyzed again. If there is still blurriness and unclearness, the corresponding area is marked as the lesion area, and image processing techniques are associated to more accurately analyze the features in the image.

[0045] Further, to train a high-performance CNN model, usually thousands to tens of thousands of labeled training images are required. These images need to accurately indicate the targets or categories in the image so that the model can learn features and make effective predictions. Therefore, by fine-tuning the screening parameter, the proportion of the second standard line area, on a small target data set to adapt to the process of adjusting the selection processing mode of the selection processing unit, the model conducts refined learning for the target task to improve its performance on the specific data set, enhancing the applicability of the model. At the same time, it also reduces the dependence on large-scale labeled data and saves the computing resources required for training the deep learning model. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic structural diagram of the rare disease recognition system based on neural network according to an embodiment of the present invention;

[0047] Figure 2 It is a schematic structural diagram of the sample screening module according to an embodiment of the present invention;

[0048] Figure 3 It is a schematic structural diagram of the sample scanning module according to an embodiment of the present invention;

[0049] Figure 4 It is a schematic structural diagram of the analysis unit according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0050] In order to make the objectives and advantages of the present invention clearer, the present invention will be further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0051] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present invention and do not limit the protection scope of the present invention.

[0052] It should be noted that in the description of the present invention, the terms indicating directions or positional relationships such as "upper", "lower", "left", "right", "inner", "outer", etc. are based on the directions or positional relationships shown in the drawings. This is only for convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention.

[0053] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and defined, the terms "installation", "connection", and "connection" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0054] Please refer to Figure 1 as shown, which is a schematic structural diagram of a rare disease recognition system based on a neural network according to an embodiment of the present invention. The present invention provides a rare disease recognition system based on a neural network, including,

[0055] a sample collection module, which is connected to an external big data platform and is used to collect rare disease samples at a preset collection period to obtain an original data set;

[0056] a sample screening module, which preprocesses and screens the original data set in sequence to obtain a target data set;

[0057] a sample scanning module, which acquires each scanned image in the target data set to identify and segment the corresponding lesion area in each scanned image to obtain a roughly segmented image;

[0058] a sample recognition module, which recognizes the roughly segmented image, extracts image features, compares the image features with lesion features, and marks the matching roughly segmented image as a target image sample; a sample evaluation module, which calculates the percentage of the number of target image samples correctly predicted by the model in the total number of image samples to obtain a real-time accuracy rate to determine whether to adjust the screening parameters.

[0059] The screening parameter is the proportion of the second standard line area.

[0060] In this embodiment, through the automatic classification and recognition of lung images, a rare disease recognition result is output, and the rare disease recognition result includes normal tissues, benign nodules, malignant nodules, and non-nodule lesions.

[0061] By connecting to a big data platform, rare disease samples are automatically collected to improve the efficiency and consistency of data collection. The original dataset is preprocessed and screened through a sample screening module to improve the efficiency of data processing and ensure data quality. The sample scanning module uses image recognition technology to accurately segment the lesion area in the scanned image, providing an accurate basis for subsequent analysis. The real-time accuracy is calculated through the sample evaluation module to dynamically adjust the screening parameters according to the real-time accuracy, optimize the recognition process, and improve the accuracy of recognition.

[0062] See Figure 2 As shown, it is a schematic structural diagram of the sample screening module according to an embodiment of the present invention;

[0063] Specifically, the sample screening module includes a preprocessing unit, a primary screening unit, and an iterative screening unit. Among them,

[0064] The preprocessing unit is used to obtain the image data in the original dataset as the original image, and perform contour enhancement processing on the original image to obtain an intermediate image;

[0065] The primary screening unit performs a primary screening on each of the intermediate images to delete the intermediate images with unqualified quality, obtaining an intermediate dataset;

[0066] The iterative screening unit performs continuous extended scanning on the intermediate images in the intermediate dataset to obtain the target dataset.

[0067] The primary screening unit includes a data deletion subunit and a data enhancement subunit. Among them,

[0068] The data deletion subunit is used to determine whether to delete the intermediate image according to the slice thickness;

[0069] Among them, when the slice thickness exceeds 5 millimeters, the intermediate image is deleted;

[0070] The data enhancement subunit is used to perform contour edge marking on the edge pixel points of the picture area in the intermediate image area, and sequentially adjust each original brightness value to the corresponding corrected brightness value to perform data enhancement processing on the brightness data of the intermediate image;

[0071] By judging the slice thickness of the image, images with a slice thickness meeting the standard are automatically screened out to ensure sufficient detail capture.

[0072] Specifically, the data enhancement processing includes,

[0073] The data enhancement subunit respectively obtains the brightness values of each pixel point in the area within the contour as a number of original brightness values, calculates the average brightness within the contour according to the original brightness values, obtains the brightness values of each pixel in the area outside the contour to calculate the average brightness outside the contour, subtracts the average brightness within the contour from the average brightness outside the contour to obtain the real-time brightness difference, and successively adds each original brightness value to the real-time brightness difference to obtain the corresponding corrected brightness value.

[0074] By performing data enhancement processing on the brightness values of pixel points, analyzing data accurately and reliably, equalizing and adjusting the brightness values of each pixel point in the area within the contour in the image to be close to the brightness values of each pixel point in the area outside the contour of the image area, making the contour edge more obvious.

[0075] The iterative screening unit includes an image cropping subunit and a screening subunit, wherein,

[0076] The image cropping subunit is used to rotate, translate and scale the intermediate image;

[0077] The screening subunit is used to perform continuous extended scanning on the intermediate images in the intermediate dataset to delete duplicate intermediate images;

[0078] In this embodiment, by calculating the hash value of each intermediate image, through continuous extended scanning, to calculate the hash value of each intermediate image and compare it with the hash values of other intermediate images. If the Hamming distance between the two hash values is less than the threshold, it is considered that the two images are duplicates. Among the images with the same hash value, one is retained and the others are deleted.

[0079] Refer to Figure 3 as shown, which is a schematic structural diagram of the sample scanning module of the embodiment of the present invention;

[0080] Specifically, the sample scanning module includes a window sliding unit, an analysis unit, a selection processing unit and a segmentation unit, wherein,

[0081] The window sliding unit can obtain the total number of pixels of any scanned image, calculate the preset sliding window size according to the sliding scanning coefficient and the total number of pixels, divide the scanned image into several areas to be scanned with the preset sliding window size, and perform sliding scanning on the areas to be scanned;

[0082] The analysis unit is used to obtain the marked line pixels in the area to be scanned, and calculate the unit line pixel ratio and the real-time line area ratio according to the marked line pixels;

[0083] The selection processing unit is configured to determine the real-time line area ratio based on the first standard line area ratio and the second standard line area ratio, select a processing mode according to the judgment result, and perform enhancement processing on the area to be scanned according to the processing mode to obtain an image to be analyzed;

[0084] The segmentation unit is configured to segment the image to be analyzed and the area to be scanned to obtain a roughly segmented image.

[0085] In this embodiment, the preset sliding window size is the product of the scanning coefficient and the total number of pixels. The image scanning coefficient represents a set decimal value, which is related to the size and quality requirements of the image to be processed. Generally, it is ensured that the set value makes the preset scanning window size the square of an integer value, that is, the preset sliding window size can be 3×3 and can be set between 0.01 - 0.1; the sliding window is rectangular; dividing the scanned image into several areas to be scanned means that according to the current sliding window, the current scanning area is simulatedly enclosed, and the corresponding image area is accurately obtained for determination to ensure the smooth and effective progress of the determination;

[0086] By setting the scanning coefficient, the corresponding preset sliding window size can be calculated according to the actual size of the image to be processed, improving the applicability to image processing.

[0087] Refer to Figure 4 as shown, which is a schematic structural diagram of the analysis unit of the embodiment of the present invention;

[0088] Specifically, the analysis unit includes a determination subunit, a marking subunit, and a calculation subunit, where,

[0089] The determination subunit is configured to compare the ratio of unit line pixels to the ratio of standard line pixels;

[0090] The marking subunit is configured to mark the image area corresponding to the current sliding window as a line area when the ratio of unit line pixels is greater than or equal to the ratio of standard line pixels;

[0091] Wherein, when the ratio of unit line pixels is less than the ratio of standard line pixels, the image area corresponding to the current sliding window is not marked;

[0092] The calculation subunit is configured to obtain the percentage of the number of line areas in the total number of areas to be scanned to obtain the real-time line area ratio.

[0093] In this embodiment, the set ratio of standard line pixels represents the ratio of white line pixels to all pixels in the image area corresponding to the current sliding window, representing the amount of white line pixels. Generally, the ratio of standard line pixels is set to 5% - 15%, and preferably, it is set to 10%.

[0094] Specifically, the selection processing unit includes a comparison sub-unit and a processing sub-unit, where,

[0095] The comparison sub-unit is used to compare the real-time line area ratio according to the first standard line area ratio and the second standard line area ratio;

[0096] The processing sub-unit is used to perform enhancement processing on the line contour edges in the picture area when the real-time line area ratio is between the first standard line area ratio and the second standard line area ratio, so as to obtain the image to be analyzed.

[0097] Performing enhancement processing on the line contour edges includes,

[0098] Performing radial gradient transition processing on each gradient transition layer of the line contour edge;

[0099] Among them, determining the number of radial gradient transition layers includes obtaining the line contour edge segments as the initial radial gradient transition layers, obtaining each pixel point that constitutes the line contour edge segments as each edge center point, respectively calculating the brightness differences between the adjacent upper and lower layer pixel points and the edge center point starting from each edge center point, obtaining a number of real-time brightness differences, if there is a real-time brightness difference greater than the preset brightness difference, then selecting the layer where the pixel point is located as the radial gradient transition layer, and accumulating the initial number of radial gradient transition layers;

[0100] The radial gradient transition processing is to adjust the color brightness of each gradient transition layer through a superposition algorithm.

[0101] In this embodiment, the superposition algorithm adopts the superposition blending mode algorithm, which adjusts the color of the upper layer according to the color brightness of one layer, that is, if the color brightness of the current layer is less than 0.5, then multiply the color of the upper layer by 2, so as to brighten the color of the upper layer, if the color brightness of the current layer is greater than or equal to 0.5, then multiply the color of the upper layer by the sum of the two brightnesses minus the product of the two brightnesses, so as to darken the color of the upper layer, and calculate layer by layer to achieve the gradient effect.

[0102] When it is determined that the real-time line area ratio is between the first standard line area ratio and the second standard line area ratio, there may be some line edges that are blurred. Since the blurred or unclear edges are also related to the lesion, first perform enhancement processing on the line contour edges to ensure the clarity of line recognition, avoid misrecognition. After the enhancement processing, analyze the line edge pixels again. If there is still blurring and unclarity, mark the corresponding area as the lesion area and associate image processing techniques to more accurately analyze the features in the image.

[0103] Specifically, the segmentation unit includes a first segmentation sub-unit, a second segmentation sub-unit and a third segmentation sub-unit, where,

[0104] When the proportion of the real-time line area is less than the proportion of the first standard line area, the first segmentation subunit performs nodule feature recognition and segmentation on the area to be scanned according to a convolutional neural network, and obtains a rough segmentation image;

[0105] Among them, the first segmentation subunit identifies nodule features according to a convolutional neural network, segments the area where the nodule features are located, and obtains a rough segmentation image;

[0106] The second segmentation subunit identifies the contour line segments with unclear edges in the image to be analyzed, and uses the corresponding area as the rough segmentation image;

[0107] The third segmentation subunit is used to perform recognition and segmentation on the area to be scanned according to a convolutional neural network when the proportion of the real-time line area is greater than the proportion of the second standard line area, so as to identify non-nodular pulmonary lesion areas and obtain a rough segmentation image.

[0108] In this embodiment, a model based on CNN is used as a classifier. Using the method of transfer learning, a model pre-trained on ImageNet is used as the initial parameters, and then fine-tuning is performed on the pulmonary image dataset; the nodule features are round, quasi-round, with lobulated or spiculated edges, and there are cavities inside; the non-nodular pulmonary lesion area is a quasi-round white lesion with gas and liquid planes inside.

[0109] In this embodiment, the proportion of the standard line area set is used to represent the proportion of the area where the white lines appear in the image. The set proportion of the first standard line area is close to zero, and the set proportion of the second standard line area is 50%. When the proportion of the real-time line area is less than the proportion of the first standard line area, it means that nodules, fat or other features block the lines. In this case, it is necessary to analyze the shape features of the nodules according to the convolutional neural network model to analyze whether the nodules are malignant or benign, so as to classify the scanned images. When the proportion of the real-time line area is greater than the proportion of the second standard line area, it indicates that most of the white lines are clearly visible and there is no obvious nodule occlusion, but it is still necessary to further analyze whether there are non-nodular lesions through the convolutional neural network to improve the accuracy of recognition.

[0110] The sample evaluation module includes a comparison unit and a correction unit, where,

[0111] An accuracy threshold is set in the comparison unit. The comparison unit compares the real-time accuracy with the accuracy threshold to obtain a comparison result;

[0112] The correction unit obtains the comparison result,

[0113] If the real-time accuracy rate is less than the accuracy rate threshold, the proportion of the second standard line area is corrected to the proportion of the second corrected line area, and the next sample evaluation is continued with the proportion of the second corrected line area;

[0114] If the real-time accuracy rate is greater than or equal to the accuracy rate threshold, the screening parameters are not adjusted;

[0115] Among them, Nb2’ = Nb2 × [1 + (Qb - Qs) / Qb], where Nb2’ is the proportion of the second corrected line area, Nb2 is the proportion of the second standard line area, Qs is the real-time accuracy rate, and Qb is the accuracy rate threshold.

[0116] In this embodiment, the accuracy rate threshold is set to 86%. The higher the accuracy rate, the more accurate the model is.

[0117] To train a high-performance CNN model, usually thousands to tens of thousands of labeled training images are required. These images need to accurately mark the targets or categories in the images so that the model can learn features and make effective predictions. Therefore, by fine-tuning the screening parameter, the proportion of the second standard line area, on a small target data set to adapt to the process of adjusting the selection processing mode of the selection processing unit, the model conducts refined learning for the target task to improve its performance on the specific data set, enhance the applicability of the model, and at the same time reduce the dependence on large-scale labeled data and save the computing resources required for training the deep learning model.

[0118] So far, the technical solution of the present invention has been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the protection scope of the present invention.

[0119] The above are only the preferred embodiments of the present invention and are not used to limit the present invention; for those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A rare disease recognition system based on a neural network, characterized in that, Including, a sample collection module, which is connected to an external big data platform and is used to collect rare disease samples at a preset collection period to obtain an original data set; a sample screening module, which preprocesses and screens the original data set in sequence to obtain a target data set; a sample scanning module, which is used to obtain each scanned image in the target data set, determine the real-time line area ratio according to the first standard line area ratio and the second standard line area ratio, select a processing mode according to the judgment result, and identify and segment the corresponding lesion areas in each scanned image according to the processing mode and a convolutional neural network to obtain a roughly segmented image; the sample scanning module includes a window sliding unit, an analysis unit, a selection processing unit, and a segmentation unit, where the window sliding unit can obtain the total number of pixels of any scanned image, calculate a preset sliding window size according to a sliding scanning coefficient and the total number of pixels, divide the scanned image into several areas to be scanned with the preset sliding window size, and perform sliding scanning on the areas to be scanned; the analysis unit is used to obtain the marked line pixels in the area to be scanned and calculate the unit line pixel ratio and the real-time line area ratio according to the marked line pixels; the selection processing unit is used to select a processing mode according to the judgment result, and identify and segment the corresponding lesion areas in each scanned image according to each processing mode and a convolutional neural network to obtain a roughly segmented image; the segmentation unit is used to segment the image to be analyzed and the area to be scanned to obtain a roughly segmented image; the selection processing unit includes a comparison subunit and a processing subunit, where the comparison subunit is used to compare the real-time line area ratio with the first standard line area ratio and the second standard line area ratio; the processing subunit is used to enhance the line contour edge of the area to be scanned when the real-time line area ratio is between the first standard line area ratio and the second standard line area ratio to obtain an image to be analyzed; the segmentation unit includes a first segmentation subunit, a second segmentation subunit, and a third segmentation subunit, where when the real-time line area ratio is less than the first standard line area ratio, the first segmentation subunit performs nodule feature recognition and segmentation on the area to be scanned according to a convolutional neural network to obtain a roughly segmented image; wherein, the first segmentation subunit identifies nodule features according to a convolutional neural network and segments the area where the nodule features are located to obtain a roughly segmented image; the second segmentation subunit is used to identify the image to be analyzed to obtain a roughly segmented image; when the real-time line area ratio is greater than the second standard line area ratio, the third segmentation subunit performs recognition and segmentation on the area to be scanned according to a convolutional neural network to identify non-nodular pulmonary lesion areas and obtain a roughly segmented image; a sample recognition module, which recognizes the roughly segmented image, extracts image features, compares the image features with lesion features, and marks the matching roughly segmented image as a target image sample; The sample evaluation module calculates the percentage of the number of target image samples correctly predicted by the model in the total number of image samples to obtain the real-time accuracy rate, so as to determine whether to adjust the screening parameters; Among them, the screening parameter is the proportion of the second standard line area.

2. The rare disease identification system based on a neural network according to claim 1, wherein The sample screening module includes a preprocessing unit, a primary screening unit, and an iterative screening unit. Among them, The preprocessing unit is used to obtain the image data in the original dataset as the original image, and perform contour enhancement processing on the original image to obtain an intermediate image; The primary screening unit performs a primary screening on each of the intermediate images to delete the intermediate images with unqualified quality, and obtains an intermediate dataset; The iterative screening unit performs continuous extended scanning on the intermediate images in the intermediate dataset to obtain the target dataset.

3. The rare disease identification system based on a neural network according to claim 1, wherein The analysis unit includes a determination subunit, a marking subunit, and a calculation subunit. Among them, The determination subunit is used to compare the proportion of unit line pixels with the proportion of standard line pixels; The marking subunit is used to mark the image area corresponding to the current sliding window as a line area when the proportion of unit line pixels is greater than or equal to the proportion of standard line pixels; The calculation subunit is used to obtain the percentage of the number of line areas in the total number of areas to be scanned, and obtain the real-time line area proportion.

4. The rare disease recognition system based on a neural network according to claim 1, characterized in that, The processing subunit performs radial gradient transition processing on each gradient transition layer at the edge of the line contour to strengthen the edge of the line contour.

5. The rare disease recognition system based on a neural network according to claim 2, wherein The primary screening unit includes a data deletion subunit and a data enhancement subunit. Among them, The data deletion subunit is used to determine whether to delete the intermediate image according to the slice thickness; The data enhancement subunit is used to mark the edge pixel points of the picture area in the intermediate image area, and sequentially adjust each original brightness value to the corresponding corrected brightness value to perform data enhancement processing on the brightness data of the intermediate image.

6. The rare disease recognition system based on a neural network according to claim 5, characterized in that, The data enhancement subunit respectively obtains the brightness values of each pixel point in the area inside the contour as a number of original brightness values, calculates the average brightness inside the contour according to each original brightness value, obtains the brightness values of each pixel outside the contour and calculates the average brightness outside the contour, subtracts the average brightness inside the contour from the average brightness outside the contour to obtain the real-time brightness difference, and sequentially adds each original brightness value to the real-time brightness difference to obtain the corresponding corrected brightness value, so as to perform data enhancement processing on the brightness data of the intermediate image.

7. The rare disease identification system based on a neural network according to claim 1, characterized in that, The sample evaluation module includes a comparison unit and a correction unit. Among them, An accuracy threshold is set in the comparison unit. The comparison unit compares the real-time accuracy rate with the accuracy threshold to obtain a comparison result; The correction unit obtains the comparison result, and when the real-time accuracy rate is less than the accuracy threshold, corrects the proportion of the second standard line area to the proportion of the second corrected line area.

Images